You open the invoice portal, sign in through a separate window and view a document inside your accounting dashboard. It feels like one task. To the browser, it involves several relationships: a connection to a website, one page displayed inside another, and two windows that may need to communicate.
Your browser is the program displaying the pages on your computer. Each service has a server, a computer answering requests over the internet. The browser can display one page, show an additional page in a smaller area, or open a separate window. Page programs, called scripts, can also ask for more material or coordinate an intended sign-in flow. So “a website is open” can describe several connected activities, not one indivisible screen. 25
For example, you open the portal, choose an invoice, and use a trusted external service to sign in. The portal may open another window for that sign-in. Separately, the invoice viewer may appear inside the accounting dashboard. None of these arrangements is automatically unsafe. Their connections do explain why protection needs to target a specific unwanted behaviour, and why a change can affect normal work.
Security settings can place limits on each relationship. Understanding the difference helps explain why “turn on every strict option” is not a complete plan. A useful protection must block the unwanted behaviour without accidentally preventing the work the service is meant to do. 5 8 13 14
Remembering to use the secure connection
A web address can begin with HTTP or HTTPS. HTTPS protects the connection with encryption. A site may redirect a visitor from HTTP to HTTPS, effectively replying, “Use the secure address instead.” But that reply and a browser that already knows to insist on HTTPS are not the same thing. 5
The difference appears at the first step. If the browser first contacts the unencrypted address, that initial exchange lacks the protection of HTTPS. The server’s normal answer may be a redirect, but somebody able to interfere with the unprotected exchange could try to change what happens before the browser reaches the secure page. A rule already stored in the browser removes that dependence for later covered requests: the browser changes the request to HTTPS before sending it. 5
A website can teach the browser a rule: “For this name, use HTTPS for the period I specify.” The mechanism is called HTTP Strict Transport Security, or HSTS. The browser must receive a valid instruction over a secure connection; an instruction delivered over plain HTTP is not accepted for this purpose. On later requests covered by the remembered rule, it uses HTTPS without first depending on an insecure redirect. 5
This is a remembered instruction, not a repair service. It does not repair a fault in the secure connection or decide who may read an invoice. If a covered site’s secure connection fails, the browser does not simply fall back to insecure access. That is part of the protection, but it also means the website operator has to keep HTTPS working. 5
The rule can extend to names below the main domain: for example, a customer portal or a support site under the same business domain. These are called subdomains. Extending HSTS to them should follow an understanding of which services exist and whether they all support HTTPS. Another contractor may operate one of them. 5
There is also a way for browsers to know the rule before a first visit: a built-in list, called an HSTS preload list. This is a separate commitment. Adding the word preload to an instruction does not prove that the domain has reached browsers’ lists, and removal is not immediate. Chromium’s submission guidance recommends HSTS but warns against treating preloading as a general default. A missing preload setting is not automatically a defect. 15
Deciding who may put your page inside theirs
A dashboard can show another web page inside a rectangular area of its own page. This is called a frame. It can be useful: an accounting application might deliberately embed a document viewer. But a deceptive page might also try to place a sensitive screen inside misleading surroundings so that someone clicks a control without understanding what it does. This kind of deception is called clickjacking. 8
Imagine the deceptive page advertises a harmless “Show discount” button while positioning a real sensitive control from an embedded screen where your click lands. The problem is not that you guessed a password incorrectly. You may be using your own legitimate sign-in, but you are misled about what you are pressing. A browser that refuses the unauthorized embedding prevents that presentation of the protected screen. It does not check every message or every design on the internet for deception. 8
The page being displayed can tell the browser who is allowed to contain it. This uses Content Security Policy: the page’s instructions limiting what the browser may load, run or display. Within those instructions, the rule called frame-ancestors controls who may embed this page. In ordinary language: “May anyone embed this screen, only our own application, certain partners, or nobody?” The correct answer depends on the intended service. 8
Notice the direction. We are asking who may place the invoice page inside their page, not which other pages the invoice page may display. Those are different permissions. Also, having some other CSP rule does not automatically supply this framing restriction. 8
For a portal intended to work inside a partner’s dashboard, denying every form of embedding could break a legitimate function. For a screen that should never be embedded, an explicit restriction may fit well. Neither choice verifies the amount on an invoice or the employee’s authority to approve it; it governs how the screen can be presented.
Separating windows and controlling outside content
Now return to the sign-in window. A page and the window it opens can have a relationship that the application relies on. A setting called Cross-Origin-Opener-Policy, or COOP, can separate certain window relationships. Under relevant COOP rules, links between a window and the window that opened it can be severed. 13
That can provide useful separation, but an external sign-in or payment window may need an intended relationship to remain available. The developer has to choose and test the appropriate behaviour. The protection concerns those window relationships, not everything that can happen in a browser. 13
Why separate the windows at all? Their connection can give the pages ways to affect or observe one another. The browser can cut the relevant connection under COOP rules. It does not have to close either window to do that. The exact result depends on the chosen rule and both pages; an intended sign-in flow still needs checking. 13
A similarly named setting, Cross-Origin-Embedder-Policy, or COEP, concerns certain content loaded from elsewhere. Depending on the chosen mode, it may require appropriate permission to use that content, or omit credentials such as cookies on certain requests. The modes are different; simply saying “isolation enabled” does not explain which rules apply. 14
For example, an externally supplied image may normally load without that supplier explicitly agreeing to its use in this page. One COEP arrangement requires appropriate permission for such outside material; another can allow certain loads but remove cookies, the small stored values that may identify a visitor to the outside service. The latter therefore requests the material without that sign-in information rather than simply accepting every request unchanged. Together with an appropriate window policy, these rules support stronger separation. They are not a general-purpose malware scan, and an image available only to a signed-in user might stop loading when the credentials are omitted. 14
You do not need to configure these settings yourself. Keep the three purposes separate: HSTS concerns using the secure connection, framing rules concern who can contain the page, and the isolation settings concern window relationships and outside content. If a proposed change breaks sign-in or document viewing, ask the developer to explain that particular relationship. That gives you a concrete discussion about protection and functionality, rather than a contest over who can enable the most switches.
Technical extra: what window separation changes
The reason for separation is that window relationships provide some ways for pages to affect or observe each other. Even when ordinary browser rules already restrict access to another site’s content, retaining a connection between windows can expose information through more indirect behaviour. COOP can move the documents into separate browser groups and sever the relevant opener relationship, reducing those opportunities. “Separate” here does not mean that one window disappears or that all communication everywhere becomes impossible. The exact policy and both pages’ circumstances matter. 13
Sources
[5] IETF: RFC 6797 HTTP Strict Transport Security
[13] MDN: Cross-Origin-Opener-Policy
[14] MDN: Cross-Origin-Embedder-Policy